Machine learning method and information processing apparatus

Abstract

A non-transitory computer-readable recording medium stores a program for causing a computer to execute a process, the process includes acquiring training data that includes a first sentence expressed in a first language and a second sentence expressed in a second language, identifying a named entity and parts of speech from the first sentence, and generating, based on the training data, a translation model that includes an attention mechanism for the named entity and the parts of speech.

What is claimed is: 
     
       1. A non-transitory computer-readable recording medium storing a program for causing a computer to execute a process, the process comprising:
 acquiring training data that includes a first sentence expressed in a first language and a second sentence expressed in a second language; 
 identifying a named entity and parts of speech from the first sentence; and 
 generating, based on the training data, a translation model that includes a joint attention mechanism for the named entity and the parts of speech. 
 
     
     
       2. The non-transitory computer-readable recording medium according to  claim 1 , wherein
 the translation model includes an encoder and a decoder, 
 a calculation result of the encoder is input to the decoder, 
 the decoder includes the joint attention mechanism, and 
 the process further comprises: 
 converting, based on a result of the identifying the named entity, the first sentence into named entity identifying information in which each word included in the first sentence is converted to either a symbol that represents named entities or a symbol that does not represent named entities, and 
 training the translation model by using a value obtained by inputting a first calculation result and a second calculation result to the joint attention mechanism, the first calculation result being calculated as a result of inputting the first sentence to the encoder, the second calculation result being calculated as a result of inputting the named entity identifying information to the encoder. 
 
     
     
       3. The non-transitory computer-readable recording medium according to  claim 2 , the process further comprising:
 converting, based on a result of the identifying the parts of speech, the first sentence into part-of-speech identifying information in which each word included in the first sentence is converted to a symbol that corresponds to a part of speech of the word; and 
 training the translation model by further using a value obtained by inputting a third calculation result to the joint attention mechanism, the third calculation result being calculated as a result of inputting the part-of-speech identifying information to the encoder. 
 
     
     
       4. The non-transitory computer-readable recording medium according to  claim 1 , wherein the joint attention mechanism includes a combination of an attention mechanism for the parts of speech and an attention mechanism for the named entity. 
     
     
       5. A machine learning method, comprising:
 acquiring, by a computer, training data that includes a first sentence expressed in a first language and a second sentence expressed in a second language; 
 identifying a named entity and parts of speech from the first sentence; and 
 generating, based on the training data, a translation model that includes a joint attention mechanism for the named entity and the parts of speech. 
 
     
     
       6. The machine learning method according to  claim 5 , wherein
 the translation model includes an encoder and a decoder, 
 a calculation result of the encoder is input to the decoder, 
 the decoder includes the joint attention mechanism, and 
 the machine learning method further comprises: 
 converting, based on a result of the identifying the named entity, the first sentence into named entity identifying information in which each word included in the first sentence is converted to either a symbol that represents named entities or a symbol that does not represent named entities, and 
 training the translation model by using a value obtained by inputting a first calculation result and a second calculation result to the joint attention mechanism, the first calculation result being calculated as a result of inputting the first sentence to the encoder, the second calculation result being calculated as a result of inputting the named entity identifying information to the encoder. 
 
     
     
       7. The machine learning method according to  claim 6 , further comprising:
 converting, based on a result of the identifying the parts of speech, the first sentence into part-of-speech identifying information in which each word included in the first sentence is converted to a symbol that corresponds to a part of speech of the word; and 
 training the translation model by further using a value obtained by inputting a third calculation result to the joint attention mechanism, the third calculation result being calculated as a result of inputting the part-of-speech identifying information to the encoder. 
 
     
     
       8. The machine learning method according to  claim 5 , wherein the joint attention mechanism is based on a combination of an attention mechanism for the parts of speech and an attention mechanism for the named entity. 
     
     
       9. An information processing apparatus, comprising:
 a memory; and 
 a processor coupled to the memory and the processor configured to: 
 acquire a first sentence expressed in a first language; 
 identify a named entity and parts of speech from the first sentence; and 
 generate a second sentence expressed in a second language, based on the first sentence and a translation model that includes a joint attention mechanism for the named entity and the parts of speech. 
 
     
     
       10. The information processing apparatus according to  claim 9 , wherein
 the translation model includes an encoder and a decoder, 
 a calculation result of the encoder is input to the decoder, 
 the decoder includes the joint attention mechanism, and 
 the processor is further configured to: 
 convert, based on a result of the identification of the named entity, the first sentence into named entity identifying information in which each word included in the first sentence is converted to either a symbol that represents named entities or a symbol that does not represent named entities, and 
 generate the second sentence by using a value obtained by inputting a first calculation result and a second calculation result to the joint attention mechanism, the first calculation result being calculated as a result of inputting the first sentence to the encoder, the second calculation result being calculated as a result of inputting the named entity identifying information to the encoder. 
 
     
     
       11. The information processing apparatus according to  claim 10 , wherein
 the processor is further configured to: 
 convert, based on a result of the identification of the parts of speech, the first sentence into part-of-speech identifying information in which each word included in the first sentence is converted to a symbol that corresponds to a part of speech of the word; and 
 generate the second sentence by further using a value obtained by inputting a third calculation result to the joint attention mechanism, the third calculation result being calculated as a result of inputting the part-of-speech identifying information to the encoder. 
 
     
     
       12. The information processing apparatus according to  claim 9 , wherein the joint attention mechanism includes a combination of an attention mechanism for the parts of speech and an attention mechanism for the named entity.